cerebras.modelzoo.data.common.h5_map_dataset.dataset.MultimodalSimpleHDF5Dataset

class cerebras.modelzoo.data.common.h5_map_dataset.dataset.MultimodalSimpleHDF5Dataset(*args, **kwargs)

Bases: cerebras.modelzoo.data.common.h5_map_dataset.dataset.MultiModalHDF5Dataset

Dataset class to handle image preprocessing in multimodal datasets.

This class is largely the same as the parent class MultimodalHDF5Dataset except with added support for multiple images and intermingling of text and images.

Parameters

params (dict) – A dictionary containing parameters that HDF5Dataset accepts along with the following add-ons: - “img_data_dir” (str): the path to the directory containing the images. - “image_data_size” (list[int]): the final C x H x W shape of the image. - “transforms” (list[dict]): a specification of the torchvision transforms.

Methods

generate_sample
load_state_dict
map
preprocess_img
state_dict

Attributes

by_sample

previous

cerebras.modelzoo.data.common.h5_map_dataset.dataset.MultiModalHDF5Dataset

next

cerebras.modelzoo.data.common.h5_map_dataset.preprocess_pile